analyzed 17 experiments (conducted between 2022 and 2025) that compared the creativity, originality, and usefulness of ideas generated by AI models (such as ChatGPT, Claude, and DALL·E) compared to those ideas of humans.
The study found no evidence that these AI models outperform humans in generating creative, original, or useful ideas. Also, newer AI models such as GPT-4 did not produce more creative ideas than earlier versions. This is the first meta-study to challenge a growing assumption that AI has surpassed human creativity.
Generative AI models are trained on large dataset collections of human-made texts, images, and other media. As a result, when prompting them for creative ideas, they often generate what the model has learned to be the “most likely creative idea.” This limits its potential for generating genuinely creative and original outcomes.
The experiments that were analyzed in this study used clear and language-based tasks, tasks that play to AI’s strengths. But these tasks do not reflect testing creativity on truly complex, unclear, and novel problems, problems that often characterize real-world creative practice in art and design. We expect creative thinking in the latter to be much harder with AI.
Creative idea generation is shaped by lived experience, ethical context, and shared human meaning, to name a few crucial factors that are hard, if not impossible, to simulate meaningfully with AI. The study shows how, despite recent claims, AI surpassing human creativity remains a distant prospect.
An article detailing the study has been accepted for publication and presentation at the European Conference on Cognitive Ergonomics, taking place on 7–10 October 2025 in Tallinn, Estonia, and will also be presented at the 9th MIC Conference 2025: Founding Creativity Studies, held on 3–6 September 2025 in Bolzano, Italy.
In the meantime, you can access the article freely here: https://doi.org/10.31234/osf.io/9u2ke_v1
‘‘Understanding how creativity and imagination emerge from interactions with our environment will lead to improved innovation processes, tools and technologies.’’
Alwin de Rooij is Associate Professor in Situated Art, Design, and Technology at the Avans Centre of Applied Research for Art, Design and Technology (CARADT), and Assistant Professor in Creativity Research in the department of Communication and Cognition, Tilburg University.
Living in cities developed around data and acting within the inscrutable structure of our techno-society demands art and design that can help understand how we relate to these rapidly changing surroundings and to reflect on that relationship. The research group Situated Art, Design and Technology responds to this exigency by fostering a situated turn in art and design through a diverse portfolio of interdisciplinary research projects in partnership with academic and cultural partners, as well as with government and industry.
Read more
Thank you for your subscription! Please check your email inbox to confirm.